01Hybrid retrieval combining full-text search (FTS5) and vector search, fused via Reciprocal Rank Fusion, then re-ranked with temporal decay and importance boost.
02Automatic entity extraction and relation tracking to build a temporal knowledge graph, boosting top search results by graph proximity.
03Multi-machine sync via Google Drive with JSONL-based merge sync, and end-to-end encrypted passport sync using AES-256-GCM and Argon2id for secure cross-device bootstrap.
04LLM-driven compression of older memories and LLM summarization of related memories for STM-to-LTM consolidation, achieving significant token reduction.
05Zero-config local setup using Fastretrieval's built-in Qwen3 ONNX embedding and reranking models, requiring no external API keys for basic operation.
